CHAPTER-3: LOAD-FLOW FOR

UNBALANCED RADIAL DISTRIBUTION

NETWORKS

3.1 Introduction

In the power system planning and operational studies load-flow analysis plays a key role.
Definite applications, mainly in distribution automation and optimization of a power
system, necessitate repetitive load-flow solution and in these applications, it is very
significant to solve the load-flow problem as proficiently as possible. Nowadays, as
distribution networks becoming more complex and these networks need for effective and
consistent system operation. Therefore, the utmost significant system exploration tool,
power- flow studies, should have the ability to handle different system arrangements with
adequate precision and execution time. In various cases, it is perceived that the radial
distribution networks are unbalanced because of single-phase (1-Φ), two-phase (2-Φ) and
three-phase (3-Φ) loads. Thus, power-flow solution of any unbalanced radial distribution
network, requires distinctive treatment. Moreover, the methods for power-flow analysis of
unbalanced distribution networks cannot be formulated by merely lengthening the single-
phase balanced techniques to three phase systems. A three-phase load-flow method has to
report problems like modelling of various forms of component connections, defining initial
point for three-phase load-flow results, as there are phase changes and transformation ratios
for every phase and at diverse buses.

3.2 Load-flow study of unbalanced radial distribution networks

The load-flow analysis is the elementary computation utilized to analyze a certain power
system functioning at steady-state beneath the itemized conditions of power input, load
demand, and network arrangement. The load-flow solution of any radial distribution
network gives information on the voltage magnitudes and angles at each bus, the real and

Page | 22
reactive power supplied at the substation, the real and reactive power-flows in each line
section, and the losses occurring in the system.

The outcomes of a power-flow exploration can be used for functioning purposes to

compute different operational states of an existing network. These can also be utilized in
the forecasting phases to estimate conceivable future schemes. In network reconfiguration,
the load-flow analysis is cast-off to evaluate the overall real power loss for a given system
arrangement in a directive to validate it alongside other arrangements. Applications like
the integration of DG or capacitor need repetitive power-flow solution. The outcomes of
the power-flow are also used in the estimation of the electrical load, and functioning
limitations.

The development of digital computers and their wide use in the power system made to
develop many algorithms in 1950’s. The most popular algorithms such as Gauss-Seidel,
Newton-Raphson and their decoupled versions have been designed for transmission
systems. These methods are unsuitable for distribution systems due to its own inherit
characteristics like radial in structure, an unbalanced distributed load, a large number of
nodes and branches and wide range of resistance and reactance values. The basic building
block of the distribution system is shown in Figure 3.1.

Figure 3.1: Basic block diagram of distribution system

Page | 23
The high R/X ratio of distribution system causes the system to be ill-conditioned for
conventional power-flow methods, especially the fast-decoupled newton method, which
diverges in most cases. Therefore, due to the characteristics of radial distribution networks,
conventional methods for the study of the radial distribution networks should be modified.

3.3 Load-flow analysis of unbalanced radial distribution networks

In this work, a novel method on load-flow analysis of the unbalanced radial distribution
networks is developed, which includes assessments of an algebraic expression of the
magnitude of the voltage and no trigonometric functions as opposite to the normal
load-flow analysis. The detailed literature survey has already been discussed in Section 2.2
(Chapter 2). In all the discussed work on load-flow, majority of the research was based on
the balanced distribution network. Here the load-flow technique for practically existing
networks, which are unbalanced, is presented. Computationally the proposed load-flow is
more efficient and it needs less computer memory storage because all the data can be stored
in the vector form. The electrical equivalent of a radial distribution system is shown in
Figure 3.2.

Figure 3.2: Electrical equivalent of the distribution network

The voltages, real power and reactive power-flow at the buses of the unbalanced
distribution network are computed by using subsequent recursive equations. The n th line is
represented by a series impedance and the impedance of the phases is given below.

Page | 24
 Z An  RAn  jX An (3.1)

Z Bn  RBn  jX Bn (3.2)

ZCn  RCn  jX Cn (3.3)

where RAn, RBn, RCn are the resistances of phase A, B, and C at the nth line, X An, XBn, XCn
are the reactances of phase A, B, and C at the nth line. The real power demand is P D and
reactive power demand is QD, which is related to complex power demand (SD). The
complex power for phases A, B and C are represented by the Equation (3.4), Equation (3.5)
and Equation (3.6) respectively.
 S D  A( k )   PD  A( k )  j  QD  A( k ) (3.4)

 S D B ( k )   PD  B ( k )  j  QD  B ( k ) (3.5)

 S D C ( k )   PD C ( k )  j  QD C ( k ) (3.6)

The branch current flowing from the bus k to the bus k+1 (branch-jj) and the phases A, B
and C are given by the Equation (3.7), Equation (3.8) and Equation (3.9) respectively.
VA(k) δ A(k)  VA(k 1 ) δ A(k 1 )
. I A(jj)  (3.7)
Z A(jj)

VB ( k )  B ( k )  VB ( k 1)  B ( k 1)
I B ( jj )  (3.8)
Z B(jj)

VC ( k )  C ( k )  VC ( k 1)  C ( k 1)
I C ( jj )  (3.9)
Z C(jj)

Where VA ( k )  A ( k ) , VB ( k )  B ( k ) , VC ( k )  C ( k ) are the voltage magnitude and phase angle

of the phase A, phase B and phase C at the kth node,

VA ( k1)  A ( k1) , VB ( k 1)  B ( k 1) , VC ( k 1)  C ( k 1) are the voltage magnitude and phase angle

of the phase A, phase B and phase C at (k + 1)th node. Here current through the branch jj
is entering at its receiving-end node k + 1, which is the sum of the load current of node
(k + 1) and the current flowing through the other branches emanating from this node.

Page | 25
The branch currents of the phase A, phase B and phase C can also be represented by
Equation (3.10), Equation (3.11) and Equation (3.12) respectively.
PA( k 1)  jQA( k 1)
I A ( jj )  (3.10)
VA*( k 1)

PB ( k 1)  jQB ( k 1)

I B ( jj )  (3.11)
VB*( k 1)

PC ( k 1)  jQC ( k 1)

I C ( jj )  (3.12)
VC*( k 1)

In Equation (3.10), Equation (3.11) and Equation (3.12) the power P m (k+1) and Qm(k+1) are
the total real and reactive power load at the (k+1)th node respectively and are expressed by
nbus nbranch
Pm ( k 1)    PD m( j )    PLoss m( jj )
j  k 1 jj  jj 1
(3.13)

nbus nbranch
Qm ( k 1)    QD m( j )    QLoss m( jj )
j  k 1 jj  jj 1
(3.14)

In Equation (3.13) and Equation (3.14), the phases A, B and C have been denoted as ‘m’
for simplification, jj = 1, 2, … , nbranch where nbranch = nbus  1. Branch jj is the branch
between the nodes k and k+1.
For end - node
Pm n bus
 
 PDm nbus
(3.15)

From Equation (3.7) and Equation (3.10), the following relation can be expressed.
Vm ( k )  m ( k )  Vm ( k 1)  m ( k 1) Pm ( k 1)  jQm ( k 1)
 (3.16)
Z m ( jj ) Vm*( k 1)

Vm ( k )  m ( k )  Vm ( k 1)  m ( k 1) Pm ( k 1)  jQm ( k 1)

i.e., 
Z m ( jj ) Vm ( k 1)    m ( k 1)

Vm ( k ) Vm ( k 1)   m ( k )   m ( k 1)   Vm ( k 1)   Pm ( k 1)  jQm ( k 1)  ( Z m ( jj ) )

2
i.e., (3.17)

Vm ( k ) Vm ( k 1) cos  m ( k )   m ( k 1)   Vm ( k 1)  j Vm ( k ) Vm ( k 1) sin  m ( k )   m ( k 1) 

2

i.e., (3.18)
  Pm ( k 1) Rm ( jj )  Qm ( k 1) X m ( jj )   j  Pm ( k 1) X m ( jj )  Qm ( k 1) Rm ( jj ) 

Separating real and imaginary parts of Equation (3.18), Equation (3.19) and Equation
(3.20) are obtained.

Page | 26
 Vm ( k ) Vm ( k 1) cos  m ( k )   m ( k 1)   Vm ( k 1)  Pm ( k 1) Rm ( jj )  Qm ( k 1) X m( jj )
2
(3.19)

Vm ( k ) Vm ( k 1) sin  m ( k )   m ( k 1)   Pm ( k 1) X m ( jj )  Qm ( k 1) Rm ( jj ) (3.20)

Squaring Equation (3.19) and Equation (3.20) and adding, we have

2
Vm ( k ) Vm ( k 1)   Vm ( k 1)   Pm ( k 1) Rm ( jj )  Qm ( k 1) X m ( jj )     Pm ( k 1) X m ( jj )  Qm ( k 1) Rm ( jj ) 
2 2 2 2

 
(3.21)
On expanding and rearranging Equation (3.21),
 Vm ( k )
2

 2.0   Pm ( k 1) Rm ( jj )  Qm ( k 1) X m ( jj )     Vm ( k 1)   Rm2 ( jj )  X m2 ( jj )  Pm2( k 1)  Qm2 ( k 1)   0
4 2
Vm ( k 1)
 2 
 
(3.22)
The solution of Equation (3.22) is given by

 2 2
 
 P V 
R
 m( k 1) m( jj )  Q X
m ( k 1) m ( jj ) 
m( k )

2


  R 2
m( jj )  X m ( jj )  m ( k 1)
2
P 2
 Qm ( k 1)  
2

Vm( k 1)     (3.23)

 Vm( k ) 
2


 Pm( k 1) Rm( jj )  Qm( k 1) X m( jj )  
 2 
 
The voltage of each node of the phases A, B, and C can be computed by Equation (3.23).
From Equation (3.19) and Equation (3.20), we have
 Pm ( k 1) X m ( jj )  Qm ( k 1) Rm ( jj ) 
 m ( k )   m ( k 1)  tan  1 
V 2

 m ( k 1)  Pm ( k 1) Rm ( jj )  Qm ( k 1) X m ( jj ) 
 Pm ( k 1) X m ( jj )  Qm ( k 1) Rm ( jj ) 
 m ( k 1)   m ( k )  tan 1   (3.24)
V 2
 Pm ( k 1) Rm ( jj )  Qm ( k 1) X m ( jj ) 
 m ( k 1)
The voltage angle of each node can be computed from Equation (3.24).
Similarly, the real and reactive power losses at branch jj in the three phases are given by
Equation (3.25) and Equation (3.26) respectively.
 P2 
m ( k 1)  Qm ( k 1)
2

 Loss m( jj ) m ( jj )
P  R    (3.25)
 V
2

 m ( k 1) 

Page | 27
  P2  Qm2 ( k 1) 
 QLoss m( jj )  X m ( jj )   m ( k 1)
 (3.26)
 Vm ( k 1)
2

 
Here, the substation or the root node voltage value is taken as 1 p.u. and the remaining
node voltages and their angles are computed by using Equation (3.23) and Equation (3.24)
respectively, where m denotes the phases A, B and C.
In this load-flow analysis of distribution network, further, the impact of distribution
transformer has been included. Studies found that while concluding the total losses of an
unbalanced distribution network, transformer model should be included as this is having a
high impact on large-scale unbalanced distribution system problems such as load-flow and
system losses. Chen et al. (1991b) in their work had shown the transformer modelling, core
loss calculations and admittance matrix formation of transformer models to be utilized in
various studies and their suggested transformer model is shown in Figure 3.3.

Figure 3.3: Transformer model

The core losses of the transformers can be expressed by Equation (3.27) and Equation
(3.28) respectively.
Sn
 
2
2 CV
Pp .u .  A V  Be (3.27)
Sb

Sn
 
2
2 FV
Q p .u .  D V  Ee (3.28)
Sb

In Equation (3.27) and Equation (3.28), V is the bus voltage in per unit, S n is nominal
transformer capacity in kVA, Sb is base power in kVA. The constant values are

Page | 28
“A=0.00267, B=0.734×10-9, C=13.5, D=0.00167, E=0.268×10-13, F=22.7” Here A, B, C,
D, E and F are constants, which are machine dependent.
The following steps are used to solve the power-flow of unbalanced distribution networks
by the suggested method.
Step-1 : Start
Step-2 : Read line data and load data, the maximum number of iterations, base kV
and base MVA and the value of 
Step-3 : Make V(k) = 1 + j0 for all nodes in each phase
Step-4 : Make initial loss PLOSS = 0 and QLOSS = 0
Step-5 : Step Iteration Count (ITC) = 1
Step-6 : Compute load current of each node for each phase
Step-7 : Compute the flow of current through each branch
Step-8 : Compute voltage of each node using Equation (3.23) and voltage angle
of each node using Equation (3.24)
Step-9 : Compute V(k)=  V(k) OLD  V(k) NEW 
Step-10 : Compute V(k)MAX
Step-11 : If V(k)MAX < , GOTO Step-14
Step-12 : Increase ITC by 1
Step-13 : If (IT  ITMAX) GOTO Step-6. Otherwise, Print Solution Has Not
Converged and GOTO Step-15
Step-14 : Compute real and reactive power losses of each branch of the system
using Equation (3.25) and Equation (3.26) respectively. Display bus
voltage, ITC, PLOSS and QLOSS.
Step-15 : Stop

3.4 Simulation results and discussions

The proposed method for load-flow is implemented on MATLAB 2013 working platform
(Windows 8, Intel Core i3-3210, 3.19 GHz). The parameters used for simulation are shown
in Table 3.1.

Page | 29
 Table 3.1: The parameters used for Simulation
Parameters Specifications
Substation voltage 1 0 p.u.
Maximum branch voltage 1.05
Minimum branch voltage 0.95
Maximum branch current 1.0

3.4.1 Load-flow results of 19-node unbalanced distribution network

The network data is available in Vulasala et al. (2009). Figure 3.4 shows the 19-node
unbalanced radial distribution network having 11 kV and 1 MVA as base values. The
detailed data of this system has been presented in Appendix-A. The magnitude of voltage
in p.u. of each node and voltage angle (deg.) shown in Table 3.2. The real power loss,
reactive power loss, minimum voltage of each phase with status of open switches is shown
in Table 3.3.

Figure 3.4: 19-node unbalanced distribution network

Page | 30
Table 3.2: Magnitude of voltage in p.u. of each node and voltage angle (deg.) of 19-
node unbalanced distribution network

Node Voltage (p.u.) Voltage angle (deg.)

Number Phase A Phase B Phase C Phase A Phase B Phase C
1 1.0000 1.0000 1.0000 0.00 -120.00 120.00

2 0.9875 0.9891 0.9880 0.01 -119.98 120.05

3 0.9854 0.9887 0.9863 0.00 -119.98 120.06

4 0.9824 0.9839 0.9830 0.03 -119.97 120.06

5 0.9820 0.9837 0.9828 0.03 -119.97 120.07

6 0.9793 0.9808 0.9801 0.04 -119.96 120.07

7 0.9786 0.9803 0.9796 0.04 -119.96 120.08

8 0.9728 0.9738 0.9735 0.06 -119.94 120.08

9 0.9659 0.9660 0.9657 0.08 -119.91 120.09

10 0.9563 0.9555 0.9550 0.09 -119.86 120.09

11 0.9550 0.9543 0.9533 0.10 -119.86 120.10

12 0.9548 0.9538 0.9536 0.11 -119.87 120.10

13 0.9544 0.9534 0.9521 0.10 -119.85 120.11

14 0.9545 0.9539 0.9528 0.10 -119.86 120.11

15 0.9527 0.9512 0.9513 0.11 -119.83 120.12

16 0.9534 0.9515 0.9522 0.13 -119.86 120.10

17 0.9537 0.9534 0.9523 0.10 -119.90 120.11

18 0.9538 0.9532 0.9521 0.10 -119.82 120.10

19 0.9514 0.9494 0.9501 0.13 -119.86 120.10

Page | 31
 Table 3.3: Load-flow results of 19-node unbalanced distribution network

Parameters Phase A Phase B Phase C

Real Power Loss (kW) 4.85 4.56 4.78

Reactive Power Loss (kVAr) 1.99 1.90 1.97

Magnitude of Minimum Voltage (p.u.) 0.9514 (19) 0.9494 (19) 0.9501 (19)
Switches Opened 19, 20

3.4.2 Load-flow results for 25-node unbalanced distribution network

A 4.16 kV and 30 MVA, 25-node unbalanced distribution network is taken as a common
test case. It has 3 tie switches and its total load is 3.239 + j2.393 MVA. Network data are
available in Vulasala et al.(2009). The detailed data of this system has been presented in
Appendix-B. The 25-node unbalanced test system is shown in Figure 3.5. The magnitude
of voltage in p.u. of each node and voltage angle (deg.) for each phase is shown in Table
3.4. The real power loss, reactive power loss and minimum voltage of each phase with
status of open switches is shown in Table 3.5.

Figure 3.5: 25-node unbalanced distribution network

Page | 32
Table 3.4: Magnitude of voltage in p.u. of each node and voltage angle in deg. of
25-node unbalanced distribution network

Node Voltage (p.u.) Voltage angle (deg.)

Number Phase A Phase B Phase C Phase A Phase B Phase C

1 1.0000 1.0000 1.0000 0.00 -120.00 120.00

2 0.9702 0.9711 0.9755 -0.57 -120.41 119.31
3 0.9632 0.9644 0.9698 -0.70 -120.52 119.15
4 0.9598 0.9613 0.9674 -0.77 -120.57 119.08
5 0.9587 0.9603 0.9664 -0.76 -120.57 119.08
6 0.9550 0.9559 0.9615 -0.56 -120.36 119.29
7 0.9419 0.9428 0.9492 -0.55 -120.30 119.27
8 0.9529 0.9538 0.9596 -0.56 -120.35 119.29
9 0.9359 0.9367 0.9438 -0.55 -120.28 119.26
10 0.9315 0.9319 0.9395 -0.55 -120.26 119.25
11 0.9294 0.9296 0.9376 -0.55 -120.26 119.25
12 0.9285 0.9287 0.9369 -0.55 -120.25 119.26
13 0.9287 0.9287 0.9373 -0.55 -120.26 119.25
14 0.9359 0.9370 0.9434 -0.55 -120.27 119.26
15 0.9338 0.9349 0.9414 -0.55 -120.27 119.25
16 0.9408 0.9418 0.9483 -0.55 -120.30 119.27
17 0.9347 0.9360 0.9420 -0.55 -120.27 119.26
18 0.9573 0.9586 0.9643 -0.70 -120.50 119.15
19 0.9524 0.9544 0.9600 -0.69 -120.49 119.16
20 0.9548 0.9563 0.9620 -0.70 -120.49 119.15
21 0.9537 0.9549 0.9605 -0.69 -120.49 119.16
22 0.9518 0.9525 0.9585 -0.69 -120.48 119.17
23 0.9565 0.9584 0.9648 -0.76 -120.57 119.08
24 0.9544 0.9565 0.9631 -0.76 -120.57 119.07
25 0.952 0.9547 0.9612 -0.76 -120.57 119.08

Page | 33
 Table 3.5: Load-flow results of unbalanced 25-node unbalanced distribution network

Parameters Phase A Phase B Phase C

Real Power Loss (kW) 52.92 54.44 43.97
Reactive Power Loss (kVAr) 60.43 52.32 55.87

Magnitude of Minimum Voltage (p.u.) 0.9285 (12) 0.9287 (12) 0.9369 (12)
Switches Opened 25, 26, 27

3.4.3 Performance comparison

The available method presented by Chen and Yang (2010) is simulated again on the present
platform and implemented on unbalanced distribution networks. The proposed load-flow
method is also implemented for the above two unbalanced distribution networks. The
comparison of the proposed method and the existing method proposed by Chen and Yang
(2010) in terms of the relative CPU time and iteration number is shown in Table 3.6 for the
unbalanced distribution networks.

Table 3.6: Comparison of relative CPU Time and Iteration Number of the proposed
method with method (Chen and Yang (2010))

Relative CPU Time Iteration Number

Unbalanced
Distribution
Proposed Method (Chen Proposed Method (Chen and
Networks
Method and Yang (2010)) Method Yang (2010))

19-node Network 1 1.13 5 6

25-node Network 1 1.18 6 7

3.5 Conclusion
An efficient load-flow method has been proposed in this chapter. The proposed load-flow
analysis for unbalanced radial distribution networks takes less CPU time for all cases and
takes the least number of iterations in some cases compared to the existing method given
by Chen and Yang (2010). CPU time spent in computing and identifying the set of possible
solutions depends on the time necessary for the objective functions evaluation, time for
verification of defined constraints by power-flow computing and active losses.

Page | 34